Finance - Accounting Seminars

 

 

Truth-telling is essential to efficient allocation of resources but is often undermined by strategic misreporting. We show how verifiably fair randomness, complemented by other blockchain functionalities, can lead to truth-telling at the cost of introduction of random noise in information transmission. The optimal trade-off between truthfulness and precision yields equilibria that outperform those attainable without randomization in information transmission. We characterize two classes of information equilibria: information designated to directly trigger value transfer via decentralized consensus (“enforcive information”), and strategic information sharing between Bayesian-rational decision makers (“advisive information”). Our findings demonstrate usefulness of randomized transmission in broad financial settings and illustrate the benefits of blockchain technology for optimal mechanism design.

Using a simple model of firm borrowing with standard ingredients, we show that commonly used empirical approaches in the literature do not recover the impact of credit supply shocks on loan-level lending, on total firm-level borrowing or on real outcomes. We propose new estimators that recover these effects. We apply our methodology to the 2011 credit crisis in Spain and show that it implies significantly smaller effects of loan supply shocks than those generated by current empirical approaches.

Abadie et al. (QJE 2023) propose two new variance estimators, one analytical and one bootstrap, for the important use case of a binary treatment where the number of clusters is finite (e.g., U.S. states), there are many treated and control units in each cluster, and only some clusters are observed. They study a setting where treatment is randomly assigned within cluster but the treatment effect and probability of assignment to treatment vary across clusters. Their simulations show that when all clusters are observed, conventional clustered standard errors (c.s.e.’s) can be severely conservative, but they find correct s.e.’s and coverage for their proposed two-stage cluster bootstrap (TSCB), with both OLS and cluster fixed effects, and good performance for an analytical variance estimator with cluster fixed effects. We confirm that c.s.e.’s are conservative when all clusters are observed, but show that their results are a corner case. When less than all clusters are observed (even slightly less), the TSCB produces neither correct s.e.’s nor correct coverage and can be inferior to c.s.e.’s or the wild cluster bootstrap. Their analytical variance estimator is also unreliable when only some clusters are observed. Which variance estimator comes closer to correct coverage depends on the sample, the treatment, the outcome, and the number of observed clusters. Moreover, a conservative s.e. does not necessarily imply conservative coverage. Including versus omitting even a single influential cluster (California, in their dataset) can greatly affect s.e. and coverage estimates

TBD

TBD

TBD
Transaction Fee Mechanisms Robust to Welfare-Increasing Collusion

Barriers to Collusion-resistant Transaction Fee Mechanisms

 

 

TBD

TBD